data sharing opportunities in childhood, adolescent and ... · bloomberg distinguished professor of...
TRANSCRIPT
Data Sharing Opportunities in Childhood, Adolescent and Young Adult (AYA) Cancer
Research for the National Cancer Institute
Report of the Board of Scientific Advisors on the Childhood Cancer Data Initiative (CCDI)
June 15, 2020
NATIONAL INSTITUTES OF HEALTH National Cancer Institute
Board of Scientific Advisors
Ad Hoc Working Group in Support of the Childhood Cancer Data Initiative
CO-CHAIR
Kevin M. Shannon, M.D.
American Cancer Society Research Professor,
Roma and Marvin Auerback Distinguished Professor of Molecular Oncology
Professor, Department of Pediatrics
School of Medicine
University of California, San Francisco
San Francisco, California
CO-CHAIR
Otis W. Brawley, M.D., M.A.C.P., F.A.S.C.O., F.A.C.E.
Bloomberg Distinguished Professor of Oncology and Epidemiology
The Sidney Kimmel Comprehensive Cancer Center
Johns Hopkins University
Baltimore, Maryland
MEMBERS
Peter C. Adamson, M.D.
Global Head
Oncology Development and Pediatric Innovation
Sanofi
Cambridge, Massachusetts
Tom Curran, Ph.D.
Senior Vice President, Executive Director,
and Chief Scientific Officer
Children’s Research Institute
Donald J. Hall Eminent Scholar in
Pediatric Research
Children’s Mercy Kansas City
Kansas City, Missouri
James R. Downing, M.D.
President and Chief Executive Officer
St. Jude Children's Research Hospital
Memphis, Tennessee
Julie Guillot
Volunteer
Leukemia & Lymphoma Society
Chair, Pediatric Partnerships & Outreach
Founder, Target Pediatric AML Project
Former Tech Executive
Cisco/NetSolve, Dell, Accenture
Austin, Texas
Amanda Haddock
President
Dragon Master Foundation
Kechi, Kansas
Douglas S. Hawkins, M.D.
Chair, Children’s Oncology Group
Professor of Pediatrics
Division of Hematology/Oncology
Center for Clinical and Translational Research
Seattle Children’s Hospital
Seattle, Washington
Katherine A. Janeway, M.D.
Director of Clinical Genomics
Dana-Farber Cancer Institute
Assistant Professor of Pediatrics
Harvard Medical School
Boston, Massachusetts
Andrea Hayes-Jordan, M.D., F.A.C.S, F.A.A.P.
Byah Thompson Doxey Distinguished Professor
of Surgery
Division Chief, Pediatric Surgery
Surgeon-in-Chief
University of North Carolina Children’s Hospital
Warren Kibbe, Ph.D.
Chief
Translational Biomedical Informatics
Department of Biostatistics and Bioinformatics
Chief Data Officer
Duke Cancer Institute
Duke University School of Medicine
Durham, North Carolina
Andrew L. Kung, M.D., Ph.D.
Chairman and Professor
Department of Pediatrics
Memorial Sloan Kettering Cancer Center
New York, New York
John M. Maris, M.D.
Giulio D’Angio Chair in Neuroblastoma Research
Professor of Pediatrics
Department of Pediatrics
Children's Hospital of Philadelphia
University of Pennsylvania
Philadelphia, Pennsylvania
Samuel L. Volchenboum, M.D., Ph.D.
Associate Professor of Pediatrics
Dean, Master's Education
Associate Chief Research Informatics Officer
Associate Director, Institute for Translational
Medicine
University of Chicago
Chicago, Illinois
Ex Officio Members
Stephn J. Chanock, M.D.
Directo
Division of Cancer Epidemiology & Genetics
National Cancer Institute
National Institutes of Health
Bethesda, Maryland
Anthony Kerlavage, Ph.D.
Director
Center for Biomedical Informatics and
Information Technology
Nationl Cancer Institute
National Institutes of Health
Bethesda, Maryland
Lynne Penberthy, M.D., M.P.H.
Associate Director
Surveillance Research Program
Division of Cancer Control and Population
Sciences
Office of the Director
National Cancer Institute
National Institutes of Health
Bethesda, Maryland
Gregory H. Reaman, M.D.
Associate Director for Oncology Sciences
Office of Hematology and Oncology Products
Center for Drug Evaluation and Research
U.S. Food & Drug Administration
Silver Spring, Maryland
Malcolm Smith, M.D.
Associate Branch Chief for Pediatric Oncology
Clinica Investigations Branch
Cancer Therapy Evaluation Program
Division of Cancer Treatment and Diagnosis
National Cancer Institute
National Institutes of Health
Bethesda, Maryland
Brigitte V. Widemann, M.D.
Chief
Pediatric Oncology Branch
Center for Cancer Research
National Cancer Institute
National Institutes of Health
Bethesda, Maryland
Executive Secretary
Jaime M. Guidry Auvil, Ph.D.
Director
Office of Data Sharing
Center for Biomedical Informatics and
Information Technology
National Cancer Institute
National Institutes of Health
Bethesda, Maryland
Table of Contents
Executive Summary ...................................................................................................................... 2
Introduction ................................................................................................................................... 4
Types of Data for Collection and Aggregation ........................................................................... 6
Current Landscape of Childhood/AYA Cancer Research & Needs Analysis ......................... 9
Potential Barriers to Progress.................................................................................................... 14
Generating New Data ................................................................................................................. 16
Distinction Between Research and Clinical Data ..................................................................... 18
Engaging Diverse Array of Stakeholders for Input ................................................................. 20
Potential Opportunities for Transformative Discoveries ........................................................ 22
2 | P a g e
Executive Summary
In November 2019, Dr. Norman Sharpless charged the Board of Scientific Advisors (BSA) Ad Hoc Working
Group (WG) in Support of the Childhood Cancer Data Initiative (CCDI) to provide general guidance to the
National Cancer Institute (NCI) related to the development and implementation of the CCDI to establish
more efficient ways to share and use childhood cancer data to improve outcomes for children who develop
cancer and to better understand the biology of childhood cancers. Because childhood cancers are rare
diseases, strategies for increasing scientific understanding of childhood cancers can become models for the
entire cancer research community. Data aggregation will lead to answers that can positively affect all
children with cancer; therefore, the NCI envisions that the CCDI will develop specific and robust goals to
facilitate optimal data sharing. Efforts within the CCDI will include support for the identification of novel
therapeutic targets and approaches to advance new drug development. The WG was further charged with
how to best advance CCDI’s areas of focus and might recommend other research activities to enhance the
Initiative’s efforts.
The overall purpose of this WG report is to advise the NCI in implementing the CCDI so that this new
resource will enable broad, rapid data sharing in ways that will optimally facilitate childhood, adolescent and
young adult (AYA) cancer research and accelerate the development of better and less toxic therapies for the
benefit of pediatric and AYA cancer patients and their families. This overriding goal is foremost to the
members of the WG and we believe it should underpin all administrative and organizational decisions.
This WG report addresses the following broad thematic areas in pediatric and AYA cancer research in
individual sections: (1) Types of Data for Collection/Aggregation; (2) Landscape of Pediatric/AYA Cancer
Research Data and Needs Analysis; (3) Potential Barriers to Progress; (4) Generating New Data; (5)
Research versus Clinical Data; (6) Patients, Families, and Advocates; and (7) Potential Opportunities for
Transformative Discoveries. Specific recommendations relative to each subject area are presented in the
individual sections of this report.
We highlight here some key over-arching recommendations drawn from the more complete list presented in
the report for consideration by the NCI.
1. We recommend that the NCI commit to support ongoing data sharing over time through policy and
funding. The WG identified a lack of dedicated financial support to personnel, time and other
resources needed to implement effective data sharing as a substantial barrier to achieving the overall
goals of CCDI.
2. We recommend that the NCI perform a comprehensive inventory of data, databases and shared
research infrastructures with each assessed for quality, relevance and integration feasibility to better
understand the landscape of pediatric and AYA cancer and survivorship data.
3. We recommend that the CCDI aggregate six broad categories of data: (1) clinical, treatment, and
outcome data from clinical trials and the electronic health record (EHR); (2) molecular data including
research sequencing (i.e. genomic, epigenetic and proteomic data) and clinical molecular profiling;
(3) information regarding the availability and location of archived biospecimens, including germline
and tumor DNA; (4) longitudinal population data from patients and survivors of pediatric and AYA
cancers; (5) characteristics of cell line, patient derived xenograft (PDX), and genetically engineered
mouse (GEM) models of pediatric and AYA cancers; and (6) any existing preclinical data generated
from studies performed in these models.
3 | P a g e
4. We recommend that the NCI implement strategies to make all CCDI data as representative as
possible of the full spectrum of pediatric/AYA cancer patients in the United States (U.S.).
Recruitment targets should reflect the percentage of minorities in the U.S. population and to achieve
this, we recommend the NCI include resources for persons obtaining consent to communicate and
educate in a racially and ethnically sensitive manner.
5. We recommend convening experts to develop consensus guidelines addressing what
clinical/molecular characterization and sample archiving should be performed and when they should
be performed for each pediatric and AYA cancer diagnosis. The WG suggests that this group
consider both clinically actionable sequencing to inform diagnosis and therapy and discovery analysis
for research purposes. Finally, the NCI should consider potential ways of paying for molecular
testing that is not currently funded by insurance or other sources. Addressing these challenging issues
is central to the achieving the CCDI’s mission of learning from every child with cancer.
6. We recommend that the NCI develop a long-term strategy for tracking pediatric and AYA cancer
patients over many decades across their lifespan. During this time, they will become independent
adults and transition their lives (and care) many times. Assigning a universal, anonymized and
privacy-protected unique patient identifier to link clinical information, biospecimens, and any
molecular data over time is a particularly appealing solution to this problem that should be strongly
considered.
The report will conclude with some potential opportunities for the NCI to consider implementing through the
CCDI that the WG feels could lead to transformative discoveries in pediatric and AYA cancer research and
survivorship.
4 | P a g e
Introduction
The Childhood Cancer Data Initiative (CCDI) was announced last year during the President’s State of the
Union address as a proposed federal investment of $50 million per year, for the next 10 years, to support
cancer research in patients and survivors of childhood, adolescent and young adult (AYA) malignancies.
Given the amount of appropriated annual funds, it was decided CCDI would focus on the critical need to
collect, analyze, and share data better in order to maximize NCI’s ongoing investment in pediatric and AYA
cancers and survivorship. Tissue samples from patients with cancer in this age group are critically limited
and a valuable resource. The overall goal of the CCDI is not simply to generate more data, but to build
processes that transform data into knowledge that moves the field forward in meaningful ways. The CCDI
supports the wider pediatric cancer community’s goal of maximizing the stated goal of “learning from every
patient” so that ultimately those patients, survivors and their families can materially benefit in terms of
higher cure rates and improved long-term health outcomes.
Approximately 16,000 children and adolescents from birth to age 15 years are diagnosed with cancer
annually in the United States. This relatively small number of patients does not account for the large impact
of pediatric and AYA cancers on the expected lifespan, particularly given the substantial proportion of adult
cancers arising in individuals who are >70 years old. While 84% of children with cancer survive 5 years or
more and are cured of their primary malignancy, a substantial proportion of patients and survivors
experience significant long-term health problems as a result of the cancer itself or the treatment. Indeed,
many studies have shown that the life expectancy of childhood cancer survivors is substantially lower than
the general population, due to treatment-related co-morbidities. These findings underscore the need to
develop more effective as well as less toxic therapies for pediatric and AYA cancers. If we could reduce the
damage to the immature nervous and cardiovascular systems, and to other organs, we could reduce the
frequency and severity of adverse late health effects. It should also be noted that many pioneering cancer
discoveries such as curative combination chemotherapy, bone marrow transplantation, molecular targeted
therapies and cellular immunotherapies were first developed in the small pediatric/AYA population, but
these discoveries go on to benefit all cancer patients. Thus, investments in pediatric/AYA cancer research
and learnings from those patient communities can speed progress for cancer patients of all ages.
Recent advances in molecular medicine have identified key driver mutations in virtually every
pediatric/AYA cancer and have also identified unique biologic subtypes within different malignancies. This
is perhaps most evident in acute lymphoblastic leukemia (ALL), which is the most common pediatric cancer.
While the overall cure rate in pediatric ALL is now ~90%, deep genomic analyses have further characterized
subtypes of ALL with a high rate of relapse and poor prognosis. Importantly, genomic profiling of those
ALL subtypes also uncovered actionable drug targets in some patients who were otherwise lacking
potentially curative treatment options. Despite the high cure rates for many pediatric/AYA malignancies,
cancer remains the leading cause of disease-related death in children. Furthermore, the outcomes for children
with certain “high risk” cancers (e.g., most sarcomas and some brain tumors) have not improved
substantially over the past two decades, despite intensification in chemoradiotherapy. Indeed, the outcomes
for some of these cancers have not changed since they were first identified and classified accurately.
In comparison to adult cancers, where lifestyle and other environmental factors often contribute significantly,
pediatric/AYA cancer is far less common on a population-wide basis and inherited factors play a larger role.
In addition, pediatric and AYA patients are treated in diverse clinical settings across the nation, including
many institutions that see few cases each year. With this said, there is a remarkable history of successfully
conducting national and international collaborative clinical trials through the NCI-funded Children’s
Oncology Group (COG) and other consortia. The relatively small number of relapsed pediatric cancers is an
additional challenge that is being addressed, in part, though networks that are conducting phase I/II trials
5 | P a g e
across different tissue histologies. These efforts include the NCI Pediatric Molecular Analysis for Therapy
Choice (MATCH) precision trial and the development of focused collaborative efforts for testing new
therapies in leukemia, neuroblastoma, brain tumors, and some other pediatric/AYA cancers.
Although genome-wide analyses have elucidated the broad molecular landscape of most pediatric and AYA
cancers, there is no national “standard of care” for molecular profiling at diagnosis or relapse and no
consensus regarding the optimal type(s) of analyses that should be performed. Although it is estimated that
some form of molecular testing is currently performed on approximately half of all pediatric cancers, there is
a need for more data in this area. In addition to representing a substantial barrier to achieving the CCDI’s
stated goal of “learning from every child with cancer”, uneven access to modern molecular analyses raises
key questions about health equity and access to care.
In addressing the mission of the CCDI, the WG identified other distinctive aspects of pediatric/AYA cancers
including: (1) lack of substantial industry support for data collection and aggregation; (2) the potential value
of performing extensive preclinical testing of new agents across different models given the small number of
patients available for clinical trials; (3) the positive role of engaged and effective advocates in promoting key
initiatives such as The Childhood Cancer Survivorship, Treatment, Access & Research (STAR) Act of 2018
and Research Acceleration for Cure and Equity (RACE) for Children Act of 2017; (4) the importance of
longitudinal follow-up and the many challenges in tracking survivors of pediatric/AYA cancers; (5) the need
to educate patients and families (especially the underserved and minority) regarding the importance of
participating in the CCDI and related tissue and blood acquisition studies; and, (6) key issues around privacy,
particularly given the implications of discovering germline mutations in cancer predisposition genes for
patients, siblings, and other family members.
The WG considered the following general questions when preparing this report:
1) What specific types of data should ideally be brought together within the CCDI?
2) What data currently exists, where is it housed, and what are the best strategies for bringing it together
(landscape and needs analysis)?
3) What are the key barriers to aggregating these existing data?
4) What new data will enhance the CCDI and how might this be generated?
5) Are there specific practical considerations for encouraging or implementing robust data sharing?
6) What tools and systems need to be created and updated to utilize shared data to improve childhood
cancer outcomes?
7) What is the perspective of the families of pediatric cancer patients and how can they best contribute
to advancing the goals of the CCDI?
8) What key scientific questions can the CCDI address over the next 2-3 years?
This report is organized into seven sections that outline the consensus opinions of the WG and conclude with
one or more specific recommendations for consideration by the NCI. The intent is to assist the NCI in
implementing the CCDI so that it will rapidly advance discovery in order to accelerate the development of
better and less toxic therapies for the benefit of pediatric/AYA cancer patients and their families. This
overriding goal is foremost to the members of the WG and should underpin all administrative and
organizational decisions.
6 | P a g e
I. Types of Data for Collection or Aggregation
The WG first discussed the general types of data that would ideally be aggregated by the CCDI and
accessible through this new NCI resource. More specific topics such as the need to systematically inventory
what data currently exist and in what form(s), operational strategies for coordinating with other NCI and
non-NCI resources, and potential barriers are discussed later in this report. The WG identified three broad
categories of data for possible inclusion in the CCDI: (1) clinical, treatment, and outcome data from
individual patients; (2) the availability of biospecimens, including germline and tumor DNA and the results
of any molecular profiling; and, (3) resources for preclinical drug testing and data generated from studies
performed in cell line, patient derived xenograft (PDX), and genetically engineered mouse (GEM) models of
pediatric and AYA cancers. Integrating these categories of data will result in a robust, multi-modal dataset
that will empower both traditional research and additional artificial intelligence efforts. The linking of
molecular data to clinical outcomes data, as well as aggregation and generation of data to support synthetic
control arms, is of critical importance and is a key enabler to advancing precision medicine. A key point of
emphasis for the WG is the perceived need to implement policies and practices for updating existing CCDI
data on a regular basis.
Clinical Data. Clinical includes data collected from both patients enrolled on clinical trials, as well as non-
clinical trials data found in the electronic health record (EHR) and other systems. Given the substantial
impact of inherited cancer predisposition mutations/syndromes on the pathogenesis of pediatric and AYA
cancers, the WG suggests that the CCDI collect information regarding a history of early-onset cancers in
other family members and/or the presence of a known cancer predisposition syndrome (e.g.
neurofibromatosis type 1 (NF1) or Down syndrome) from all pediatric and AYA cancer patients. Some
germline polymorphisms and mutations interact with the genotoxic therapies used to cure a primary cancer to
cause histologically distinct second cancers. Given this, it would also be useful for researchers to have access
to a resource that aggregates data from individuals who develop therapy-induced cancers. The WG also
agreed that rich, standardized data regarding the primary cancer diagnosis, such as tumor type/sub-type,
stage, histologic grade, treatment given, response (complete remission/partial remission/primary refractory),
and outcome (cure/relapse/death from disease, death from a non-cancer cause, morbidities), and adverse
events should be available. Information regarding any salvage treatment that relapsed patients received, and
their responses, would be of great value. Given the increasing availability and ease of collection of real-
world data/evidence, the NCI might consider strategies to leverage data from wearables and sensors as well
as information collected from patients and families through standardized patient-reported outcome measures.
Finally, the NCI should consider implementing strategies for collecting data regarding social determinants of
health on pediatric/AYA cancer patients to facilitate research into how this influences treatment decisions
and outcomes. Efforts to include all minority races and cultures also must be prioritized. Relevant data
collected should ideally address all aspects of social determinants of health, including but not limited to
education, literacy, poverty race, ethnicity and socioeconomic status.
Molecular Profiling. These types of data include: (1) cytogenetic and FISH (Fluorescence In Situ
Hybridization) analysis of tumor cells; (2) immunostaining to measure the expression of individual proteins,
such as MYC in neuroblastoma; (3) transcriptome (RNA-seq) analysis and/or any targeted or genome-wide
DNA sequencing of biospecimens; (4) epigenetic and proteomic profiling of biospecimens; and (5) data
derived from the tumor microenvironment. Ideally, paired germline/tumor sequencing data will be housed in
a data commons or similar resource through the CCDI and made available to the research community by
providing access to both curated/summated and primary data (e.g. FAST-Q sequencing files). Finally, as
there will likely be scientific value in retrospectively performing additional molecular analyses on specimens
from some patients where the clinical outcome is known, the NCI should both consider how to include
7 | P a g e
information on the existence and location(s) of archived biospecimens from pediatric/AYA cancer patients in
the CCDI. In addition, this WG suggests that the NCI review current practices for collecting and cataloging
biospecimens from pediatric/AYA cancer patients and consider implementing strategies for enhancing these
key activities, particularly in the context of clinical trials.
Data from Preclinical Drug Testing. Testing new therapies for pediatric cancer is challenging due to the
relatively small number of patients with progressive disease who become eligible for clinical trials of
experimental drugs. This restricts the number of novel investigational agents that can be evaluated in
children with specific cancers and underscores the need for robust models of pediatric cancer that can be
used to identify the most promising agents and drug combinations. Furthermore, new anticancer agents are
almost invariably developed based on activity against adult tumors, so that target mutant proteins uniquely
expressed in childhood malignancies are rarely pursued. On the other hand, the initiating “driver” mutations
are now known for many childhood cancers, and the overall mutational burden is relatively low in most
childhood/AYA malignancies. These characteristics both facilitate generating accurate GEM models and
suggest that panels of characterized PDX models will capture the key molecular features of many
pediatric/AYA cancers. In addition, many drugs are being developed that target proteins encoded by genes
that are mutated in both adult and pediatric cancers such as: ALK, BRAF, FLT3, KTM2A/MLL, KRAS/NRAS,
and NTRK. Indeed, the RACE for Children Act mandates that pharmaceutical and biotechnology implement
a pediatric development plan for these shared biochemical targets. A robust infrastructure for performing
drug testing in pediatric cancer models will facilitate access to the most promising new agents.
The NCI currently supports the Pediatric Preclinical Testing Consortium (PPTC) through a cooperative
agreement mechanism to test new agents primarily in PDX models of pediatric cancers. This panel of 261
PDX models across 27 distinct childhood cancer histologies was recently fully characterized with DNA and
RNA sequencing in work that was funded by the Alex’s Lemonade Stand Foundation. The WG views the
CCDI as a key resource for aggregating existing information about the PPTC panel and other pediatric
cancer cell line, PDX and GEM models, and for providing access to the current and future results of drug
testing. In anticipation of implementing the RACE Act, the NCI is actively developing a Pediatric Preclinical
Public-Private Partnership (PPP3) as a strategy for continuing high throughput in vitro drug testing in cell
line models and to support in vivo studies in PDX and GEM models of pediatric/AYA cancers in academia
and with industry partners. The WG envisions the PPP3 as playing a key role in collecting data generated
from this effort and for making it available to the scientific community. The WG is not aware of any
centralized resource for investigators who are interested in reviewing published and unpublished preclinical
data from studies performed in GEM and PDX models of pediatric/AYA cancers and believes that the CCDI
could potentially address this important gap.
Recommendations
1. We recommend that the CCDI aggregate six broad categories of data: (1) clinical, treatment, and
outcome data from clinical trials and the electronic health record (EHR); (2) molecular data including
research sequencing (i.e. genomic, epigenetic and proteomic data) and clinical molecular profiling;
(3) information regarding the availability and location of archived biospecimens, including germline
and tumor DNA; (4) longitudinal population data from patients and survivors of pediatric and AYA
cancers; (5) characteristics of cell line, patient derived xenograft (PDX), and genetically engineered
mouse (GEM) models of pediatric and AYA cancers; and (6) any existing preclinical data generated
from studies performed in these models.
8 | P a g e
Recommendations (continued)
2. We recommend that the CCDI aggregate data from previous molecular analyses of pediatric/AYA
cancers irrespective of whether these studies were performed for diagnostic or research purposes
and/or as part of a clinical trial.
3. We recommend that the NCI implement strategies to make all CCDI data as representative as
possible of the full spectrum of pediatric/AYA cancer patients in the United States (U.S.).
Recruitment targets should reflect the percentage of minorities in the U.S. population and to achieve
this, we recommend the NCI include resources for persons obtaining consent to communicate and
educate in a racially and ethnically sensitive manner.
4. We recommend that the 10-year CCDI effort includes plans to include new data and adapt treatments,
diagnostics and prevention strategies to the knowledge gained as methods advance.
9 | P a g e
II. Current Landscape of Pediatric/AYA Cancer Research Data and Needs Analysis
Childhood cancers are fundamentally different than malignancies arising in adults. Pediatric and AYA
cancers are typically initiated by gain- or loss-of-function mutations in a key gene that alters a single lineage-
restricted protein that plays a fundamental role in the development of their cell of origin. In addition,
germline genetic variation potently influences tumor initiation, and epigenetic somatic alterations are critical
mediators in childhood cancer phenotypes. At diagnosis, targetable mutated kinase alterations are rare in
pediatric/AYA cancers, unlike tumors acquired later in life. However, genotoxic stress from chemotherapy
and/or radiation therapy dramatically influences mutation burden at relapse and can sometimes create new
therapeutic vulnerabilities. Consequently, childhood/AYA cancers must be investigated and treated as
biologically distinct from their adult counterparts.
The RACE for Children Act has the potential to significantly and favorably alter the landscape for childhood
cancer drug development by shortening the timeline between “first-in-human” (almost always adult) and
“first-in-children” clinical trials. For drugs that show some activity in adults, it takes, on average, over 6
years before they are tested in pediatric cancers. The RACE Act is designed to address this problem by
requiring that targeted drugs developed for adult cancer indications are rapidly tested in children when the
molecular target is relevant to the growth or progression of one or more pediatric cancers. Expanding the
evidence base of the relevance of specific genetic aberrations, pathway alterations, and membrane antigen
expression to cancer survival and proliferation in children is hampered by the diminished access to clinically
actionable tumor DNA sequencing, gene and protein expression analyses, and methylome sequencing.
Furthermore, the lack of a central source for aggregation and integration of extant data developed at
individual institutions for use by clinical investigators, industry sponsors, and regulatory agencies is a
substantial impediment to translation. The CCDI can support the RACE for Children Act by addressing these
needs. Expanding the evidence base for individual targets within individual pediatric cancer types and across
the spectrum of cancer diagnoses will inform clinical research strategies for investigators, alert sponsors of
the probable requirement for early consideration of pediatric development plans for new molecularly targeted
agents, and engage the FDA and other international regulatory agencies in decision-making related to their
authority in implementing the RACE ACT and other legislative initiatives.
There are several databases and repositories that include pediatric/AYA cancer data. Some of these contain
only pediatric data, while others contain data from cancer patients of all ages. Some databases primarily host
research data, while others are for clinical care. Major efforts generating data relevant for patients and
survivors of childhood/AYA cancers within NCI and associated Cancer Centers include the trans-NIH
(National Institutes of Health) Gabriella Miller Kids First Pediatric Research Program; NCI Pediatric
MATCH trial; NCI Therapeutically Applicable Research to Generate Effective Treatments (TARGET)
initiative; the Childhood Cancer Survivor Study (CCSS), the Pediatric Immunotherapy Discovery and
Development Network (PI-DDN); and the St. Jude LIFE study. In addition, most clinical trials data in the
U.S. are collected by the NCI-funded Children’s Oncology Group (COG). The data from these efforts are
accessible through multiple, disparate data repositories within and outside of NCI such as the Genomics Data
Commons (GDC), Cancer Research Data Commons (CRDC), Kids First Data Resource Center (DRC),
Pediatric Cancer Data Commons (PCDC), St. Jude Cloud, cBioPortal, and AACR Project Genomics
Evidence Neoplasia Information Exchange (GENIE). In addition, there are several private for-profit
organizations such as Tempus, Archer, and Foundation Medicine with relevant cancer data and databases.
This list is only a partial representation of data available on childhood/AYA cancer patients and survivors. It
will be critical for CCDI to consider how to aggregate or use data from each available pediatric/AYA or
survivorship data source to meet its goals.
10 | P a g e
Unfortunately, there is no registry or catalog of existing childhood/AYA cancer data or data repositories, and
there is little to no standardization among the existing databases. Among those with molecular profiling data,
there is also high variability as to what types of information from such testing is available – from targeted
somatic gene panel sequencing, to whole exome sequencing, to whole genome and transcriptome (RNA-seq)
sequencing, as well as rich epigenetic (methylation profiling, ChIP Seq, etc.) and proteomic data sets that can
inform childhood and AYA cancer research. Additionally, the critical clinical and demographic information
associated with these molecular data are sometimes insufficient to support rich analyses.
In addition to existing data resources listed, the NCI has begun to develop initial phases of a National
Childhood Cancer Registry (NCCR) through the CCDI that will provide access to clinically relevant patient
data from childhood cancers. The various databases and registries contain very different, yet complimentary,
types of data; integrating these resources has potential significant value in helping to answer key scientific
and treatment questions in childhood/AYA cancers and can also avoid some redundancies. Ideally the CCDI
will facilitate meaningful research through connection or federation of these data and repositories to enable
interoperability.
One issue with making many different data types interoperable, regardless of the repository, is the general
lack of standards employed for each dataset. The WG identified defining and promoting the use of standards
for future data collection as a key potential role of the CCDI. One model for this activity could be the
CRDC-H (common data model for data in the NCI Cancer Research Data Commons) from the NCI’s Center
for Cancer Data Harmonization. Having a clear, defined standard will help researchers in preparing data for
submission to the various data platforms and repositories and further ensure interoperability for future data
collection.
The WG discussed two ongoing initiatives with goals aligning to the CCDI. The AACR Project GENIE and
Children’s Oncology Group’s (COG) Project:EveryChild both focus on comprehensive data collection and
open sharing of pediatric/AYA cancer sequencing data. GENIE is a major international registry of real-world
data from patients of all ages formed in 2015. It involves data sharing among 19 academic institutions.1
Today it has over 80,000 clinical-grade sequencing (CLIA/ISO) records and an established legal and
technical infrastructure. The consortium has safely shared data with a 7,500-person registered user base. A
unique value of sequencing data generated in clinical laboratories at academic institutions is the ability to
perform deep clinical annotation because the sequencing data remain linked to the medical record. The
primary goal of the registry is to improve clinical decision making, particularly in patients with rare cancers
and patients who have rare variants in common cancers, by linking genotypes with patient outcomes;
secondary goals include catalyzing clinical and translational research.
Later in 2020, there is a planned release of nearly 3,600 sequenced pediatric tumors (age at sequencing ≤ 18)
from 16 participating Institutions in GENIE. To enhance the utility of deposited pediatric data, in 2018 Drs.
Katie Janeway and Andrew Kung initiated a NCI-funded project to annotate pediatric cancer records within
GENIE from the Dana-Farber Cancer Institute (DFCI) and Memorial Sloan Kettering Cancer Center
(MSKCC), with disease specific staging, risk factors and longitudinal clinical treatment and outcome data.
PRISSMM is an artificial intelligence tool. It is a phenomics data standard developed Dr. Deborah Schrag at
DFCI. It takes unstructured data from text reports in electronic health records and structures them so that
they can be readily analyzed. PRISSMM uses a standardized framework for determining outcomes from real-
world data. In lung and other adult cancers, pathology, imaging, signs and symptoms, tumor markers, and
medical provider assessments collected into the PRISSMM model have been transformed into cancer
1 https://www.aacr.org/professionals/blog/aacr-project-genie-the-sharing-economy-comes-to-the-clinic/
11 | P a g e
outcome data and found to be predictive of expected disease outcomes1. The PRISSMM framework has been
process hardened and deployed for largescale clinical annotation within GENIE, including the Biopharma
Collaborative Project (BPC), a five-year, $36 million, pre-competitive collaboration between GENIE and
nine leading pharma companies to annotate 50,000 GENIE records2. The GENIE pediatric clinical data
annotation project has developed a pediatric version of PRISSMM adding data elements appropriate for
pediatric cancers that are, when possible, harmonized with existing or emerging national standards such as
the Toronto staging guidelines.
Taken together, Project GENIE and the PRISSMM framework represents an extant infrastructure and an
emerging standard for aggregating and sharing clinically annotated cancer genomic data. With existing pilot-
scale pediatric projects establishing the feasibility of leveraging this shovel-ready infrastructure as a standard
for pediatric cancers, the overall goals of the CCDI and AACR Project GENIE are aligned.
As previously mentioned, the Children’s Oncology Group provides the largest resource of U.S. clinical trial
data in childhood and adolescent cancers, and a majority of the COG biobanking is done through a registry
called Project:EveryChild. It was launched in 2015 to replace it former registry of histology-specific
biobanking studies. Based upon 2017-2019 data, COG enrolls approximately 6,000 patients annually on
Project:EveryChild, with >70,000 specimens banked since inception. COG enrolls 2,200 patients per year on
phase III and pilot studies and 190 patients annually on phase II studies. Enrollment is dependent upon
broad institutional participation across a network of more than 220 COG sites. The top 50 enrolling COG
institutions account for 52% of all COG enrollments, with the remainder coming from the smaller 170
institutions. The COG Biorepository in Columbus, Ohio contains tissues from more than 32,000 children
with cancer and related disease and is supported by an NCI U24 grant. However, institutional reimbursement
for tissue and data submission through Project:EveryChild is supported by philanthropic funds for patients
who do not enroll on clinical trials. The COG Biorepository is further responsible for the critical function of
assigning the Unique Specimen Identifier (USI), the unique and privacy-preserving identifier that connects
data and specimens for COG patients. The WG considers childhood and AYA biobanking to be underfunded,
which restricts the potential impact of this resource on research and developing new therapies. It is important
for NCI to consider how CCDI can support, align with or utilize tools and best practices of initiatives like
GENIE and Project:EveryChild.
Aggregating sequencing data already being generated in commercial and academic medical center
laboratories and linking it to harmonized clinical data from trials and/or the medical record will be a major
step forward. However, it’s utility will be limited by the lack of systematic implementation of precision
oncology in pediatrics in the United States. Large cohorts of children with cancer in the U.S. have had
genomic characterization of their tumors by programs such as TARGET, Gabriella Miller Kids First
Pediatric Research Program, and the St. Jude—Washington University Pediatric Cancer Genome Project.
Yet, the United States has not had any coordinated effort to comprehensively characterize every childhood
cancer on a national level, unlike many European countries. Consequently, many childhood/AYA cancers
are not profiled at all. When genomic evaluation is performed for clinical or research purposes the
approaches vary significantly. As a result, the overlap in genomic events captured across different tests can
be small. A further issue is many platforms used to characterize pediatric and AYA cancers are designed
with a focus on adult malignancies missing key genes or variants in pediatric cancers. For example, private
(not previously described) gene fusions are very important molecular targets in pediatric cancers. They are
best identified with RNA sequencing approaches designed specifically for pediatric cancers that are not
2 https://www.aacr.org/about-the-aacr/newsroom/news-releases/aacr-project-genie-begins-five-year-collaborative-research-project-with-36-million-in-new-funding/
12 | P a g e
widely available. Implementing guidelines for the optimal molecular data set that appropriately characterizes
pediatric and AYA cancers would: 1) allow the CCDI to evaluate the utility of genomic datasets to be
aggregated; and 2) highlight areas where existing efforts are failing to generate adequate genomic data and
are therefore priorities for CCDI investment in data generation. There are substantial challenges inherent in
organizing and funding a national effort to sequence pediatric/AYA cancers. With this in mind, members of
the WG recommended prioritizing deep molecular analysis (whole exome or genome + transcriptome) of
germline/diagnosis/relapse “trios” and on rare pediatric cancers and on tumors where no effective front-line
therapies currently exist.
From a technology standpoint, CCDI has enormous potential to speed discovery, but it’s also a significant
“paradigm shifting” undertaking that presents a variety of challenges and complexities, such as:
• Numerous, geographically dispersed data owners with varying needs, priorities and degrees of ability
(technical capabilities and resources) to share data. Some have invested millions in data generation
and technology platforms and may be hesitant to embrace new data sharing and generation models.
• Ten-year funding window, potentially with year-to-year budget uncertainty, that will require carefully
phasing of work and integration considerations.
• Cultural and legal challenges, like academic publishing concerns and privacy issues, introduce added
complexity.
In addition to adequate financial resourcing, investments in “time and team” for CCDI solution design and
“master” planning/phasing will be critical to maximize benefits and ensure that all projects align with the big
picture. Inclusion of a broad spectrum of subject matter experts and stakeholders in CCDI visioning and
decision-making (across diseases, organizations/institutions, and roles) – led by technology design experts
with related experience – can help mitigate risks and ensure success.
Recommendations
1. We recommend that the NCI perform a comprehensive inventory of data, databases and shared
research infrastructures with each assessed for quality, relevance and integration feasibility to better
understand the landscape of pediatric and AYA cancer and survivorship data.
2. We recommend convening experts to develop consensus guidelines addressing what
clinical/molecular characterization and sample archiving should be performed and when they should
be performed for each pediatric and AYA cancer diagnosis. The WG suggests that this group
consider both clinically actionable sequencing to inform diagnosis and therapy and discovery analysis
for research purposes. Finally, the NCI should consider potential ways of paying for molecular
testing that is not currently funded by insurance or other sources. Addressing these challenging issues
is central to the achieving the CCDI’s mission of learning from every child with cancer.
3. We recommend that CCDI efforts include collection and integration of new “ideal” data types
prospectively, on every newly diagnosed patient -- offering standardized biomarker/omics testing
(biology panel) to every child and seamlessly linking their biology, clinical, and other critical data.
13 | P a g e
Recommendations (continued)
4. We recommend that the NCI consider how to collect off-trial patient clinical data, as well as data to
enable synthetic control arms to support precision match trials.
5. We recommend that the NCI review current practices for collecting and cataloging biospecimens
from pediatric/AYA cancer patients and assigning unique and privacy-preserving identifiers and
consider implementing strategies for enhancing these key activities, particularly in the context of
clinical trials.
6. We recommend developing the National Childhood Cancer Registry (NCCR) as a national resource
for aggregating and accessing high-quality curated clinical information, molecular data, and other
associated critical patient information and make every effort to avoid costly redundancies.
7. We recommend that the NCI convene a broad spectrum of subject matter experts and stakeholders
with expertise in technology, data science and disease biology and clinical care for patients and
survivors of pediatric and AYA cancers to assist with architecture design and road-mapping of the
NCCR and overall CCDI data infrastructure.
8. We recommend that the CCDI include activities focused on generating and aggregating preclinical
data and the development of infrastructure and tools that will enhance clinical translation. These
efforts would ideally focus on the FDA’s Relevant Molecular Targets List and on fulfilling the
mandates of the RACE for Children Act. The NCI should also consider provide resources to validate
potential targets in pediatric and AYA cancers to expand the reach of precision oncology.
14 | P a g e
III. Potential Barriers to Progress
While addressing the key systemic (“landscape”) questions outlined in the previous section will be key to the
success of the CCDI, members of the WG identified other barriers to data collection, curation, and
aggregation. A lack of dedicated financial support is a substantial impediment. Specifically, the NCI, other
funding agencies, institutions, and investigators invariably underestimate the personnel, time, and other
resources needed to implement effective data sharing. Furthermore, members of the WG articulated the need
for a commitment to support ongoing data sharing over time. For example, some NCI-funded Cancer Centers
decided not to apply for the P30 supplements announced earlier this year because they are for one year only.
As resources for data aggregation and sharing are finite, the WG also discussed whether the most efficient
strategy involves supporting major centers with large existing datasets and ongoing molecular profiling
programs or reaching out more broadly to collect data from as many patients as possible from across the
country. Of the first 2,300 pediatric cancer patients enrolled in the GENIE database, over 80% were from
four institutions (MSKCC, DFCI, University of California, San Francisco (UCSF), and Children’s Hospital
of Philadelphia (CHOP)). While the GENIE leadership is enthusiastic about collecting additional data from
pediatric/AYA cancers, there is no funding to support data aggregation at additional institutions.
Contributions from biopharmaceutical companies provides resources in support of GENIE’s activities to
aggregate and curate data in common adult cancers, but there is no real incentive to do this in rare cancers,
including pediatric malignancies. However, it may be possible to implement creative public-private
partnerships with other entities.
Pediatric and AYA cancer patients who receive care at the largest pediatric oncology research centers appear
to differ in important ways from the overall pediatric/AYA cancer population. For example, Dr. Douglas
Hawkins, chair of the Children’s Oncology Group noted that patients cared for at smaller institutions are
more likely to enroll on COG Phase III national clinical trials and less likely to have access to state-of-the-art
molecular profiling until they relapse and can participate in an initiative like the NCI Pediatric MATCH trial.
Finally, enabling the CCDI to access and “federate” (i.e. form into a connected infrastructure that functions
together) pediatric/AYA cancer data generated by foundations (e.g. Foundation Medicine), other funding or
commercial entities, academic centers, and health systems (e.g. Kaiser Permanente) is highly desirable but
may also prove challenging for various reasons. The WG suggests that the NCI address these issues
proactively by providing incentives and resources to individual stakeholders. On a more positive note, if the
CCDI is successful in effectively aggregating pediatric/AYA data from multiple sources, this would establish
a model for data sharing that might also be applicable to adult cancers.
Recommendations
1. We recommend that the NCI commit to support ongoing data sharing over time through policy and
funding. The WG identified a lack of dedicated financial support to personnel, time and other
resources needed to implement effective data sharing as a substantial barrier to achieving the overall
goals of CCDI.
15 | P a g e
Recommendations (continued)
2. We recommend that the NCI consider multiple, creative ways to engage with and utilize expertise
from a wide variety of stakeholders in the pediatric and AYA cancer and survivor communities. NCI
should explore public-private partnerships with other entities including large pediatric care centers,
foundations, and industry for interoperability and sustainability models for data and infrastructure.
NCI could also consider, in addition to a panel of scientific advisors, creating a panel of technology
advisors to bring valuable outside perspectives to the CCDI.
3. We recommend that the NCI consider committing resources to consent, sequence samples from, and
retain diverse populations in the CCDI database to proactively ensure that this resource accurately
reflects the diversity of the population.
16 | P a g e
IV. Generating New Data
Generating new cancer data aligns with the recommendations of the National Cancer Advisory Board
(NCAB) ad Hoc Data Science Working Group, which highlighted the need for NCI to play a leadership role
in: (1) funding the creation and sharing of cancer research data; (2) ensuring sustainability of its investment
in the creation of such data; (3) creating mechanisms for these data to be made broadly available to the
research community; (4) defining responsible data use policies and processes; and, (5) supporting the
training of the next generation of cancer data scientists. Nowhere in the NCI portfolio are these
recommendations more profound or transformative than in pediatric and AYA cancers. Likewise, three
recommendations in the Data Science WG report focused on key areas of high relevance for the CCDI: (1)
further investments in existing high-value data sets; (2) harmonization of terminology between cancer
research data and clinical care data; (3) increase the application of artificial intelligence (AI), including
machine learning and other approaches in cancer research and implement these methods in cancer care.
To achieve these goals, it is essential that datasets, methodologies, analytic pipelines, and results be rapidly
and openly shared, with consistent metadata and annotation information also openly shared. As with the NCI
Cancer Moonshot, it will be very important for the CCDI to have a clear, open data and open science policy.
The WG suggests that NCI and NIH work together to consider and implement methods for rewarding and
reinforcing good data sharing practices. Finally, to bring in the machine learning and AI communities, it is
important to have clear approval processes and diminished barriers to data access. Most computationally
trained researchers outside of medicine do not understand the regulatory space that currently limits access to
pediatric and young adult cancer datasets. Innovative IRBs (Institutional Review Boards), data platforms,
and access policies that are designed to work together to eliminate these barriers have the potential to
revolutionize the application of these techniques to childhood cancer. The CCDI is a historic opportunity to
create a true learning health system for pediatric cancer.
For the CCDI to generate the evidence needed to change outcomes for pediatric cancer, there will need to be
investments in generating new data either for new or existing projects, aligned with the NCAB ad Hoc Data
Science recommendations. This recommendation is even more compelling for children and young adult
cancers than it is for adults, as described below:
1. Pediatric cancer is infrequent compared to adult cancer. Consequently, it is very challenging to
develop data collections with sufficient numbers of subjects for comprehensive analyses, particularly
for the rare pediatric cancers.
2. Recent studies have uncovered a high frequency of germline predisposition mutations in pediatric
cancer. In contrast to adult cancer, this implies there is a significant familial risk that is important to
document and manage. These data have implications for the entire family.
3. There is a risk that databases will not reflect the diversity of the population. Underserved populations
are underrepresented in the current genomic studies of pediatric and AYA cancers. By ensuring that
every child has access to cancer genomics we can redress this imbalance. Increasing the diversity of
databases is a fundamental principle of precision medicine. Committing to resources to consent and
retain diverse populations should be considered. Specifically, since some minority cultures are
reticent to be included in “experiments” on their children, persons obtaining consent should include
minority populations to engender trust. For all potential participants, communication at the education
level of the participant and family should be highlighted.
4. Comprehensive genomic profiles identify variants that affect outcomes in a myriad of ways. For
example, pharmacogenomic analysis can change the selection of therapies or the dose of drugs given
17 | P a g e
to individuals. Pediatric cancer survivors encounter many long-term effects, including second
malignancies, that could be ameliorated with sophisticated incorporation of genomic data.
5. Many pediatric cancers - though they may share the same name with adult cancers - are biologically
distinct from the corresponding adult cancers. For example, acute myeloid leukemia (AML) in
children differs significantly from AML in older individuals, particularly in patients >60 years old.
Generation of pediatric/AYA-specific biology data, is critical to the discovery of new therapeutic
targets and treatment strategies.
For all of these reasons and as outlined above in Recommendation #2 in the Landscapes section, we
recommend that the NCI develop a plan to generate new data and adapt treatments, diagnostics and
prevention strategies, to the knowledge gained as methods advance in accordance with the advice of an
expert consensus panel convened to address this broad subject area.
Examples of new data that could fill existing gaps include:
• Expanded cases (e.g. current data may focus on high-risk cases, but sequencing of
intermediate cases or biological extremes [exceptional responders] may be beneficial)
• Expanded data types (e.g. single sequencing cell data, germline/tumor whole genome, whole
exome, methylome, proteome, phenomics including social determinants and other
patient/family data, etc.) as warranted by disease
Recommendations
1. We recommend that the NCI seek ways to support the generation of evidence needed to change
outcomes for pediatric cancer. This will require an investment in generating data through new or
existing NCI initiatives, aligned with the NCAB ad Hoc Data Science WG report.
2. We recommend that the CCDI harmonize terminologies and coding between cancer research and
clinical care data as recommended in the NCAB ad Hoc Data Science WG report.
3. We concur with applying of AI and machine learning approaches to analyze pediatric/AYA cancer
datasets as recommended in the NCAB ad Hoc Data Science WG report.
4. We recommend that the CCDI remove barriers to data access for the broader community by
supporting the creation of data resources that simplify data access, such as limited data sets, safe
harbor data sets, and synthetic data sets that can be broadly accessed.
18 | P a g e
V. Distinctions Between Research and Clinical Data
WG members articulated key differences between data collected through a clinical trial or for other research
purposes and data obtained for routine clinical care. As summarized below, access to each type will likely
require different approaches and governance.
Research and Registry Data. Some of the major childhood/AYA cancer databases are not open to most
investigators. The WG considered how to encourage institutions to participate fully in data sharing.
Incentives are likely more effective than penalties, but data sharing should be required, with clearly defined
penalties for non-compliance. Some felt the NCI should make data sharing a clear expectation of NCI
Designated Cancer Centers and that non-competing and competitive renewal funding should be contingent
on tangible support of the CCDI. In rare circumstances the NCI could also decide to withhold other
institutional funding such as T32 awards to ensure active participation in the CCDI. This would be
regrettable if it were to become necessary. It is essential that the NIH/NCI make it clear that the expectation -
from the start - will be for broad data sharing. While embargo periods on data sharing are reasonable, the
WG suggests establishing clear expectations that any data collected using NCI funds will be made available
to the research community within 6-12 months of collection. While this “stick” approach is essential, the
“carrot” in this scenario will be the systems and processes set up through the CCDI and other programs that
make it as easy as possible for researchers to “do the right thing.” The easiest course of action should be for
researchers to share their data with the broader community. While these tactics might be effective for
NIH/NCI funded databases, other strategies may be required for privately held databases and registries.
There must also be a clear set of tools and resources for researchers to leverage for data curation and
submission. This includes educational and training resources and expert “concierge” to support researchers
preparing and submitting data. The NCI’s Center for Cancer Data Harmonization (CCDH) will be an
excellent source of tools and educational and training resources to aid researchers in data harmonization and
submission. Some additional resources are emerging through data resources like the Genomic Data
Commons (GDC) and the Gabriella Miller Kids First Data Resource Center, and we support a concerted
NCI-wide effort to make processes standardized and available to all.
Clinical Data. Leveraging data from the electronic health record (EHR) will be critical for obtaining useful
data for studying pediatric cancer. Currently, most clinical data available for further research are from
completed clinical trials and are limited by the manual abstraction and curation of data and entry into
prescribed case report forms. The NCI is advised to support efforts to standardize the extraction and
reporting of data directly from the EHR. The CCDI-funded NCCR is one project that is planning on
leveraging data from the EHR, including diagnostic data and treatment and outcome information. Since EHR
vendors have not productively participated in widespread standardized data sharing, it is incumbent on
organizations like the NCI to incentivize academic medical centers to collect, standardize, and share the rich
granular and free-text data found in the EHR.
In addition to EHR data, there may be opportunities to collect real-world data/evidence, including
patient/parent-reported outcomes, data from wearables and sensors, and from other non-traditional sources.
For instance, information from a wearable could be used to monitor patient activity, sleep, and heart rate
after receiving chemotherapy. Standardized surveys can be administered directly to patients and families
through portals such as Epic’s MyChart, with the resulting data flowing directly back into the EHR for
subsequent extraction. These surveys could leverage PROMIS (Patient-Reported Outcomes Measurement
Information System) quality measures or even gather data from families about therapies delivered across
multiple institutions; information often best reported by parents. Data about financial toxicity could be
19 | P a g e
collected by leveraging geocoded information to infer patient socioeconomic status. Technology to leverage
these data is maturing, and the most common way to move data in and out of EHRs is via HL7-FHIR (Health
Level 7 - Fast Healthcare Interoperability Resources). Most healthcare institutions now have the facility to
use FHIR to import and export data. The NCI’s Center for Cancer Data Harmonization, part of the Cancer
Research Data Commons (CRDC) ecosystem, is evaluating FHIR resources and tooling as a solution for
sharing data. Further, the NCCR has asked NCI-funded cancer centers to augment their cancer registry
reporting by leveraging additional EHR data. While these pilot projects are just starting, it is likely that many
successful implementations will utilize FHIR messaging for pushing data from the EHR to the NCI data
resources (CCDH, NCCR). The CCDI should build upon the initial efforts of these NCI resources (CCDH,
NCCR) and potentially others to leverage the current work on using FHIR to standardize the movement of
data out of EHRs.
The usefulness of any data is partially contingent on being able to connect those data to other sources of
information. For instance, genomic data in TARGET is enriched by connecting those data to the clinical
information in the Pediatric Cancer Data Commons (derived from COG clinical trials). Connecting these
data relies on a privacy-preserving linked patient identifier. The COG has solved this issue by having the
Biopathology Center at Nationwide Children’s Hospital (Columbus, OH) deidentify each patient by
assigning a publicly available Unique Specimen Identifier (USI), a six-letter code unique to each patient, at
the time of biospecimen accessioning. This code can usually be publicly displayed with any data related to
that patient. Since COG acts as the honest broker (the only group knowing the association between COG IDs
and USIs), this process relies on COG for any disambiguation activities when updating patient data from
separate sources. While potentially useful across all childhood cancer, the current process is unlikely to be
scalable without additional resources available for both COG and the Biopathology Center, which
administers the USI system on behalf of COG. In addition to the USI, there are several other privacy-
protecting patient identifier systems in production. The team from the NCCR has had a formal consultant
report prepared detailing the most impactful systems (e.g., Datavant, EUPID), and we suggest that the NCI
review this report. Importantly, the WG believes that whatever system is chosen should be utilized across
different NCI programs (i.e., CCDI, NCCR, CRDC), and non-NCI efforts should be incentivized to use the
same system.
Recommendations
1. We recommend that the NIH/NCI set clear expectations regarding the need for broad data sharing
policies governing all CCDI activities. An embargo period on data sharing is reasonable, but the
expectation should be that data collected using NCI funds will be made available to the research
community within 6 – 12 months of collection.
2. We recommend that the NCI explore barcoding approaches to data and metadata from individual
patient data so that uniqueness can be assured, and further evaluate this by funding one or more
pilot projects using existing data sets.
3. We recommend that the NCI develop a long-term strategy for tracking pediatric and AYA cancer
patients over many decades across their lifespan. During this time, they will become independent
adults and transition their lives (and care) many times. Assigning a universal, anonymized and
privacy-protected unique patient identifier to link clinical information, biospecimens, and any
molecular data over time is a particularly appealing solution to this problem that should be
strongly considered.
20 | P a g e
VI. Engaging Diverse Array of Stakeholders for Input
Patient and Family Perspectives: Engaged and informed advocacy is a cornerstone of the pediatric/AYA
cancer community and played a key role in passage of critical legislation, including the STAR and RACE for
Children Acts. In addition, multiple philanthropic foundations are dedicated to curing childhood cancer and
fund essential career development and research activities across the nation.
Pediatric/AYA cancer advocates contributed valuable perspectives and suggestions to the WG. First and
foremost, the families of children with cancer strongly support the overall goal of the CCDI to create a
national data ecosystem that will advance research and treatment. In this spirit, they expressed the view that
molecular and other data generated from pediatric/AYA cancer patients belongs to the patients and their
families and not to a given institution, lab, or organization. Accordingly, the WG suggests that the NCI
explore creative ways for parents to “opt-in” to share (de-identified) data collected on their child (versus
institutional or organizational ownership of data), especially genomic data. Platforms already exist that allow
parents and patients to control use of their data in a dynamic way, and the WG suggests that the NCI review
these models. The NCI should consider developing the CCDI’s infrastructure in a way that could allow
patients and families to see their data and allow them to share it with other families, clinicians, and/or
researchers. There should also ideally be a mechanism for switching control of that consenting process from
parent to child when the child comes of age.
Traditionally, longitudinal data has been difficult to collect and even harder for families to access. As this
information is both important for research and critical for decision making around care, the WG strongly
encourages the NCI to develop the CCDI with the capacity to aggregate longitudinal follow-up data
regarding outcomes and quality of life in pediatric/AYA cancer patients. We also recommend that
authoritative, data-driven health care recommendations be collected in an accessible format for
pediatric/AYA cancer patients and their families through the CCDI. Effective linkage to the CCSS, St. Jude
Life and other cancer survivorship efforts will be essential for achieving this goal. To ensure that CCDI is
delivering desired benefits, the NCI should consider developing CCDI-specific key performance indicators
(KPIs; e.g. childhood cancer performance metrics/dashboard) and regular progress reports, like tracking
number of new targets discovered.
In addition to their involvement as parents, family members, friends, and patients, pediatric/AYA cancer
advocates possess valuable knowledge, experience and perspectives due to their various professional
backgrounds, expertise, and cancer journey. Given this, the WG suggests that the NCI explore ways of
harnessing these relevant skills to advance specific aspects of the CCDI in domains such as data architecture,
information technology planning, and legal affairs through a rigorous nomination and selection process. In
addition, the WG encourages that the NCI to consider involving pediatric/AYA advocates and industry
stakeholders as participants in the CCDI governance structure, oversight or review processes, and change
management activities (e.g. communications, process or policy changes).
And finally, siloes are a problem not unique to the research community -- it exists in the “advocacy/funding
community” as well. First, foundations inspired to support childhood cancer research have difficulty finding
each other -- a barrier to forming partnerships and collaboratively funding large-scale, strategic efforts.
Second, families and social workers experience the continued challenge of locating financial and other
support resources amid the sea of charities. Projects to enable funder visibility and partnering, as well as
enabling patients to easily locate and centrally apply for support resources could be small-scale but visible
“wins” for CCDI.
21 | P a g e
Recommendations:
1. We recommend that the NCI consider multiple, creative ways to engage with and utilize expertise
from a wide variety of stakeholders in the pediatric and AYA cancer and survivor communities.
NCI should explore public-private partnerships with other entities including large pediatric care
centers, foundations and industry for interoperability and sustainability models for data and
infrastructure. NCI could also consider, in addition to a panel of scientific advisors, creating a
panel of technology advisors to bring valuable outside perspectives to the CCDI.
2. We recommend that the NCI explore creative ways for parents to “opt-in” to share (de-identified)
data collected on their child (versus institutional or organizational ownership of data), especially
genomic data.
3. We recommend that the NCI consider developing the CCDI’s infrastructure in a way that could
allow patients and families to see their data and allow them to share it with other families,
clinicians, and/or researchers. There should also ideally be a mechanism for switching control of
that consenting process from parent to child when the child comes of age.
22 | P a g e
VII. Potential Opportunities for Transformative Discoveries
Given the unpredictable nature of scientific and clinical discovery, the WG acknowledged the inherent
challenges of identifying specific areas where the CCDI might serve a platform for catalyzing “game
changing” advances. With this said, the WG identified the following specific key areas where the NCI might
consider undertaking new initiatives:
• Developing a national strategy to offer appropriate biospecimen collection and genomic testing to
every pediatric/AYA patient with cancer within two years. While this is a large and complex topic,
specific priorities include: (1) developing a comprehensive plan for collecting and archiving
diagnostic germline and tumor samples from all pediatric/AYA cancer patients irrespective of
whether they are enrolled on a clinical trial and for linking this to treatment data and outcome; (2)
implementing a national standard for performing deep sequencing of the ~15% of pediatric/AYA
cancers that relapse after an initial response to front-line therapies (germline, diagnostic, relapse
triads); and, (3) consider what type(s) of molecular profiling should be performed at diagnosis in
different pediatric/AYA cancers to guide care and enhance discovery.
• Aggregate data from existing cell line, patient derived xenografts (PDX), and genetically engineered
mouse (GEM) models of pediatric or AYA cancers) that might inform rapid pediatric clinical
translation in accordance with the RACE for Children Act. Consider funding focused efforts to test
promising new agents in the most relevant pediatric cancer models.
• Identify patients who have a remarkable initial response to conventional chemotherapy and/or
targeted therapies. Deep molecular profiling of these “outliers” have provided mechanistic insights
and identified key dependencies in adult cancers. Systematically pursuing this approach will likely
also be informative in pediatric/AYA cancers.
• The COG Biorepository includes multiple rare childhood cancer specimens. A focused biologic
effort to delineating the molecular landscapes (and potential therapeutic vulnerabilities) of specific
rare cancers should leverage the COG Biorepository and other archived specimens. Strategies to
increase submission of high value tissues could increase the number of cases available.
• The COG Biorepository has a large number of germline samples that could be used for
predisposition gene discovery, especially for histologies for which large germline analysis has not
been done (e.g. liver tumors, germ cell tumors, etc.).
• Improve biobanking efforts at intake by adding quality control, digital image generation, and nucleic
acid extraction for all children diagnosed with specific solid tumors.